Source: Osler

Osler provides a comparative legal analysis of two divergent rulings that highlight the lack of international consensus on AI and copyright: the UK High Court's decision in Getty v. Stability AI and the Hamburg Regional Court's decision in GEMA v. OpenAI. While the UK court found that the Stable Diffusion model did not constitute an infringing copy because it did not "store" the training images, the German court reached the opposite conclusion regarding OpenAI's model. The German ruling emphasized that the model's ability to reproduce song lyrics in response to simple prompts demonstrated that the works were effectively "memorized" and thus reproduced, even if not stored as traditional file copies.

The article explains that this divergence stems from fundamental differences in copyright theory and evidence. German law views "reproduction" broadly to include any fixation perceptible to humans (even indirectly), whereas the UK court focused on the technical absence of the original image data in the model weights. Furthermore, the German court rejected OpenAI's reliance on the Text and Data Mining (TDM) exception, arguing that it does not cover the "memorization" that allows for competing exploitation of the work. This split creates a complex compliance landscape for AI developers, suggesting that a model considered lawful in London might be deemed infringing in Berlin.